FIRST LIGHT LBT AO IMAGES of HR 8799 Bcde at 1.6 and 3.3 Μm: NEW DISCREPANCIES BETWEEN YOUNG PLANETS and OLD BROWN DWARFS∗

Home , HR 8799, HR 8799 b, HR 8799 c, HR 8799 d

FIRST LIGHT LBT AO IMAGES OF HR 8799 bcde AT 1.6 AND 3.3 μm: NEW DISCREPANCIES BETWEEN YOUNG PLANETS AND OLD BROWN DWARFS∗

Andrew J. Skemer1,PhilipM.Hinz1, Simone Esposito2, Adam Burrows3, Jarron Leisenring4, Michael Skrutskie5, Silvano Desidera6,DinoMesa6, Carmelo Arcidiacono2,7, Filippo Mannucci2, Timothy J. Rodigas1, Laird Close1, Don McCarthy1, Craig Kulesa1, Guido Agapito2, Daniel Apai1,8, Javier Argomedo2, Vanessa Bailey1, Konstantina Boutsia9,10, Runa Briguglio2, Guido Brusa9, Lorenzo Busoni2, Riccardo Claudi6, Joshua Eisner1, Luca Fini2, Katherine B. Follette1, Peter Garnavich11, Raffaele Gratton6, Juan Carlos Guerra9,JohnM.Hill9, William F. Hoffmann1, Terry Jones12, Megan Krejny12, Jared Males1, Elena Masciadri2, Michael R. Meyer4, Douglas L. Miller9, Katie Morzinski1, Matthew Nelson5, Enrico Pinna2, Alfio Puglisi2, Sascha P. Quanz4, Fernando Quiros-Pacheco2, Armando Riccardi2, Paolo Stefanini2, Vidhya Vaitheeswaran1,JohnC.Wilson5, and Marco Xompero2 1 Steward Observatory, Department of Astronomy, University of Arizona, 933 N. Cherry Ave, Tucson, AZ 85721, USA 2 Istituto Nazionale di Astrofisica, Osservatorio Astrofisico di Arcetri, Largo E Fermi 5, 50125 Firenze, Italy 3 Department of Astronomy, Princeton University, 4 Ivy Lane, Princeton, NJ 08544, USA 4 Institute for Astronomy, ETH Zurich, Wolfgang-Pauli-Strasse 27, CH-8093 Zurich, Switzerland 5 Department of Astronomy, University of Virginia, 530 McCormick Road, Charlottesville, VA 22904, USA 6 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Padova, Vicolo dell’ Osservatorio 5, I-35122 Padova, Italy 7 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Bologna, Via Ranzani 1, 40127 Bologna, Italy 8 Department of Planetary Sciences, University of Arizona, 1629 E. University Blvd., Tucson, AZ 85721, USA 9 Large Binocular Telescope Observatory, University of Arizona, 933 N. Cherry Ave, Tucson, AZ 85721, USA 10 Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Roma, Via Frascati 33, 00040 Rome, Italy 11 Department of Physics, University of Notre Dame, 225 Nieuwland Science Hall, Notre Dame, IN 46556, USA 12 School of Physics and Astronomy, University of Minnesota, 116 Church Street S.E., Minneapolis, MN 55455, USA Received 2012 February 24; accepted 2012 April 20; published 2012 June 8

ABSTRACT As the only directly imaged multiple planet system, HR 8799 provides a unique opportunity to study the physical properties of several planets in parallel. In this paper, we image all four of the HR 8799 planets at H band and 3.3 μm with the new Large Binocular Telescope adaptive optics system, PISCES, and LBTI/LMIRCam. Our images offer an unprecedented view of the system, allowing us to obtain H and 3.3 μm photometry of the innermost planet (for the first time) and put strong upper limits on the presence of a hypothetical fifth companion. We find that all four planets are unexpectedly bright at 3.3 μm compared to the equilibrium chemistry models used for field brown dwarfs, which predict that planets should be faint at 3.3 μm due to CH4 opacity. We attempt to model the planets with thick-cloudy, non-equilibrium chemistry atmospheres but find that removing CH4 to fit the 3.3 μm photometry increases the predicted L (3.8 μm) flux enough that it is inconsistent with observations. In an effort to fit the spectral energy distribution of the HR 8799 planets, we construct mixtures of cloudy atmospheres, which are intended to represent planets covered by clouds of varying opacity. In this scenario, regions with low opacity look hot and bright, while regions with high opacity look faint, similar to the patchy cloud structures on Jupiter and L/T transition brown dwarfs. Our mixed-cloud models reproduce all of the available data, but self-consistent models are still necessary to demonstrate their viability. Key words: brown dwarfs – instrumentation: adaptive optics – planetary systems – planets and satellites: atmospheres – stars: individual (HR 8799) Online-only material: color figures

1. INTRODUCTION ously is particularly powerful given their connected formation histories and appearances. Efforts are underway to characterize the first generation of HR 8799 is a young, A5V star with a λ Boo deficiency of directly imaged extrasolar planets. A principal focus has been heavy metals and three distinct circumstellar dust structures the HR 8799 planetary system (Marois et al. 2008, 2010), (Marois et al. 2008; Cowley et al. 1969; Gray & Kaye 1999;Su which with four planets is currently unique as a directly im- et al. 2009). There is some disagreement about the age of the aged multiple-planet system. Studying these planets simultane- system. Traditional age-dating methods, such as galactic space motion and Hertzsprung–Russell diagram position, suggest that HR 8799 has an age of 20–160 Myr (Moor´ et al. 2006; Marois ∗ The LBT is an international collaboration among institutions in the United et al. 2008; Hinz et al. 2010; Zuckerman et al. 2011), while States, Italy, and Germany. LBT Corporation partners are as follows: The ∼ University of Arizona on behalf of the Arizona university system; Istituto astroseismology estimates are more consistent with 1Gyr, Nazionale di Astrosica, Italy; LBT Beteiligungsgesellschaft, Germany, which would make the planets significantly more massive brown representing the Max-Planck Society, the Astrophysical Institute Potsdam, and dwarfs (Moya et al. 2010). Interestingly, the dynamical stability Heidelberg University; The Ohio State University, and The Research of the planets themselves places upper limits on the masses of Corporation, on behalf of The University of Notre Dame, University of Minnesota, and University of Virginia. the planets (Fabrycky & Murray-Clay 2010; Moro-Mart´ın et al.

1 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal. 2010; Sudol & Haghighipour 2012), which directly converts ular Telescope Interferometer (LBTI). Additionally, we describe to a young system age, based on the planets’ photometry and our data reduction procedure, which is an implementation of evolutionary models (Burrows et al. 1997; Chabrier et al. 2000). the Locally Optimized Combination of Images (LOCI) algo- An important implication of the relative youth and low masses rithm. In Section 3 we estimate photometry for the four planets, of the HR 8799 planets is that their appearances and atmospheric based on our LOCI reductions. In Section 4 we use our H- properties might be different than field brown dwarfs, which can band image to search for additional companions interior to HR have the same effective temperatures as giant planets while being 8799 e, taking advantage of the unprecedented contrast afforded older and more massive. Brown dwarf spectra have been used as by the LBT AO system. In Section 5 we present thick-cloud, proxies for giant planet spectra to plan direct imaging surveys non-equilibrium chemistry model atmospheres and mixed-cloud and to interpret early discoveries. However, initial results show atmospheres in an effort to explain the appearances of the that there are several key differences between the atmospheres HR 8799 planets. We conclude in Section 6 and make sugges- of giant exoplanets and brown dwarfs. tions for future work characterizing HR 8799 and other directly For field brown dwarfs, the L→T spectral-type transition imaged exoplanets. A companion paper, Esposito et al. (2012), occurs at ∼1200–1400 K, where dust clouds settle/condense describes the instrumental setup for the AO system and PISCES below the photosphere (Saumon & Marley 2008), and CO is in detail, provides an independent analysis of the H-band data converted to CH4 (Burrows et al. 2003; Geballe et al. 2002). For (along with new Ks-band data), and presents astrometry and a the HR 8799 planets, clouds are suspended in the photosphere new orbital analysis of the system. at lower effective temperatures (900–1200 K) than is typical for brown dwarfs (Currie et al. 2011; Madhusudhan et al. 2011; 2. OBSERVATIONS Barman et al. 2011a). Additionally, there appears to be more CO 2.1. PISCES H Band than CH4 relative to equilibrium chemistry models, implying that convection is mixing hot material into the photosphere We observed HR 8799 at H band (λ = 1.66 μm; FWHM = ↔ faster than the CO CH4 reaction can re-equilibrate (Hinz et al. 0.29 μm) with PISCES (McCarthy et al. 2001) on UT 2011 2010;Barmanetal.2011a). Similar but more extreme results October 16, during Science Verification Time for the LBT FLAO have been found for 2MASS 1207 b, a 5–7 Mjup companion to system. The LBT’s FLAO system (PI: Simone Esposito) is a a25Mjup TW Hya brown dwarf primary (Chauvin et al. 2004; 672-actuator deformable secondary AO system that makes use Skemer et al. 2011;Barmanetal.2011b). Evidently, the HR of an innovative pyramid wavefront sensor, producing high- 8799 planets and 2MASS 1207 b look similar to L-type brown Strehl-ratio, low-background images over a broad wavelength dwarfs, despite having effective temperatures more consistent range (Esposito et al. 2010, 2011). At the time of our obser- with T-type brown dwarfs. vations, one AO system was installed on the right telescope, Multiwavelength photometry and spectroscopy are the keys so for the observations presented in this paper, only one 8.4 m to understanding the differences between brown dwarfs and mirror is used. PISCES (PI: Don McCarthy) is a 1–2.5 μmim- giant planets. In particular, working over a broad wavelength ager with a Hawaii 1024 × 1024 HgCdTe array, which at the range is critical for understanding clouds, chemistry, and the LBT (single 8.25 m aperture) critically samples a diffraction- radiative budget of extrasolar planets. The challenge of working limited point-spread function (PSF) at H band (with a plate scale over a broad wavelength range is that adaptive optics (AO) sys- of 0.0193 pixel−1). For the observations in this paper, PISCES tems perform better at longer wavelengths where atmospheric was installed on the front bent-Gregorian focus of the LBT’s turbulence is less severe. But background radiation increases at right side to take advantage of the LBT’s new AO system, while longer wavelengths, and most AO systems have numerous warm facility infrared cameras are being delivered and commissioned. optics, which can make working at long wavelengths imprac- During our H-band observations of HR 8799, conditions tical (Lloyd-Hart 2000). The sweet spot for most AO systems were photometric and the natural seeing, as measured by a has typically been the near-infrared (∼1–2.5 μm), although for differential image motion monitor (DIMM) on the telescope extrasolar planets, there is a benefit to working at L (3.8 μm), structure, was as good as ∼0.9. We obtained 901 images with where the planet–star contrast improves (Heinze et al. 2008). 2 s integrations, over the course of 2 hr (90◦ of sky rotation) The Large Binocular Telescope (LBT) AO system can in- with the telescope rotator turned off angular differential imaging crease the wavelength range over which we study extrasolar (ADI; Marois et al. 2006). PISCES’ readout time is 6 s, which planets. With its 672-actuator deformable secondary mirror (in- means our observations were inefficient. However, high-contrast stalled on one side of the telescope at the time of our obser- observations are usually limited by PSF stability rather than vations), the system produces unprecedented image quality and photon noise, so the long readouts had a negligible effect on contrast at short wavelengths. And because it is a deformable our final results. The 2 s integrations saturated out to a radius of secondary AO system, it has a minimal number of warm optics, ∼0.15. We were not able to obtain unsaturated H-band images so that background noise remains low at long wavelengths. of the star for photometry and astrometry with PISCES’ shortest In this paper, we present LBT First Light Adaptive Optics integration time, 0.8 s. (FLAO) images of HR 8799 at H band (1.65 μm) and 3.3 μm, Images were processed to remove cross-talk and persistence detecting the four planets (b–e) at both wavelengths. Both im- as described in McCarthy et al. (2001)usingthecorquad ages are superior to previous attempts at these wavelengths due correction software.13 We then dark subtracted, flat fielded, and to the high performance of the LBT’s AO system. Our images distortion corrected the images. Finally, images were aligned by are the first detections of HR 8799 e at H band and 3.3 μm maximizing their cross-correlation. We processed the aligned and the first unambiguous detections of HR 8799 b and d at images using the LOCI algorithm (Lafreniere` et al. 2007), which 3.3 μm. In Section 2, we give a basic description of the instru- has been shown to produce higher contrast images than other mental setup for the FLAO system, the near-infrared imager, algorithms, such as ADI. In the terminology of Lafreniere` et al. PISCES, and the mid-infrared imager, L- and M-band Infrared Camera (LMIRCam), which is a component of the Large Binoc- 13 http://aries.as.arizona.edu/∼observer/dot.corquad.pisces

2 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Figure 1. LBT First Light AO images of the HR 8799 planetary system at H band and 3.3 μm. These images constitute the first detection of HR 8799 e at either wavelength and the first unambiguous detections of HR 8799 b and d at 3.3 μm. (A color version of this figure is available in the online journal.)

(2007), we used Nδ = 1, NA = 300, g = 1, and a 1 pixel which itself contains a 2–5 μm channel (LMIRCam) and an 14 subtraction region. The values for NA and g are adopted 8–13 μm channel (Nulling Optimized Mid-Infrared Camera, from Lafreniere` et al. (2007), and we use Nδ = 1 instead of NOMIC). LMIRCam uses a Hawaii 2RG 2048 × 2048 HgCdTe Nδ = 0.5 to suppress self-subtraction, although we note that array, which oversamples λ = 2–5 μm PSFs (with a plate scale −1 changing Nδ to 0.5 has a negligible effect on our photometry. of 0.0106 pixel ) when using only one telescope primary. Our implementation of the LOCI algorithm also includes an Conditions during our 3.3 μm observations were photometric, FWHM-sized mask around every subtraction region to further with a natural seeing (as measured by the telescope DIMM) suppress self-subtraction. of ∼1.1. We obtained 1920 images with 3.8 s integrations After reducing all 901 images with LOCI, we evaluated the over the course of 3.3 hr (110◦ of sky rotation). The 3.8 s noise of the images (standard deviation of pixel count) within an exposure saturated inside of 0.06. We also obtained short annulus from 0.3to0.5. The noise dramatically increased after (0.15 s) unsaturated images for astrometry/photometry. About the 500th image, corresponding to an increase in natural seeing. 1% of the images had bad tip/tilt residuals and were removed. We reran LOCI using just the first 500 images (1000 s, 61◦ of Images were flat fielded, and globally bad pixels (defined as sky rotation), marginally improving our final image, which is pixels further than 15% from the median flat) were replaced shown in Figure 1. HR 8799 b, c, d, and e are all clearly visible. with the average of the eight nearest good pixels. The images The LOCI algorithm assumes accurate knowledge of the star’s were then nod subtracted with nods taken every 20 images. position. However, the stellar core saturated even with PISCES’ After these basic reduction steps, we found that some pixels had fastest readout. We were able to estimate the stellar centroid to changing biases over shorter timescales than the nod subtraction. within 0.5 pixels (0.01) based on the circular symmetry of the We removed these with three steps: subtracting the 3σ -clipped PSF. We then reran LOCI with a set of different stellar centroid median of each column (to remove column bias-level effects), positions composing a 1 × 1 pixel box and found no significant median-binning the data with a 2 × 2 box (since the PSF is astrometric or photometric discrepancies between the results. oversampled by a factor of three for a single-aperture telescope), and replacing isolated bad pixels (4σ from their neighboring 2.2. LBTI/LMIRCam 3.3 μm pixels) with the median of the neighboring pixels. For the remainder of this paper, LMIRCam “pixels” refer to binned We observed HR 8799 at 3.3 μm(λ = 3.31 μm; pixels, which have a plate scale of 0.0212. FWHM = 0.40 μm) with LBTI/LMIRCam (Hinz et al. 2008; Images were processed with LOCI, using the same parameters Skrutskie et al. 2010) on UT 2011 November 16. Although and implementation described in Section 2.1. We evaluated the LBTI will eventually be used to combine the light from both noise inside a 0.2–0.4 annulus in each LOCI-processed image LBT apertures, it was used in single-aperture mode for these and removed ∼10% of the images, which had high noise, due to observations, since only one AO system was operational. LBTI residual bad pixels and/or sub-optimum AO performance. Our (PI: Phil Hinz) consists of a beam combiner (Universal Beam final image, which is shown in Figure 1, clearly shows planets Combiner), which combines the light from the two telescope ter- b, c, d, and e. tiary mirrors, and a science camera (Nulling Infrared Camera),

14 The general idea of LOCI with ADI is as follows. The stellar PSF is 3. PHOTOMETRY removed in a set of subtraction regions, which together compose the image. For each subtraction region, the stellar PSF is estimated by constructing a Since LOCI self-subtracts, we calibrated our photometry by linear combination of individual exposures that minimizes noise within a subtracting fake planets from the raw data at the positions of corresponding optimization region, which is centered on the subtraction region the detected planets and rerunning the full LOCI pipeline. Best- but is much larger. The optimization region is an annulus with an area of NA ﬁt photometry and error bars were determined by adjusting PSF cores and a ratio between its radial extent and azimuthal extent of g. Images whose parallactic angle differs by less than an amount such that a the ﬂuxes of the fake planets to determine the range of values source would move by Nδ FWHM are excluded from the optimization. resulting in reasonable subtractions. For the PISCES H-band

3 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Table 1 LBT Photometry of the HR 8799 Planets

Planet ΔH Mag with HR 8799 b Absolute H Mag Δ3.3 μm Mag with HR 8799 Absolute 3.3 μmMag HR 8799 b 15.08 ± 0.13a 10.97 ± 0.10 13.22 ± 0.11 HR 8799 c −0.90 ± 0.05 14.18 ± 0.14 9.97 ± 0.10 12.22 ± 0.11 HR 8799 d −0.85 ± 0.2 14.23 ± 0.2 9.77 ± 0.10 12.02 ± 0.11 HR 8799 e −1.2 ± 0.2 13.88 ± 0.2 9.87 ± 0.20 12.02 ± 0.21

Notes. a HR 8799 b absolute H-band photometry from Metchev et al. (2009). Absolute H-band photometry for the other planets is with respect to HR 8799 b. data, the star was saturated, so we calculated photometry for 4. CONSTRAINTS ON A HYPOTHETICAL FIFTH planets c, d, and e with respect to HR 8799 b and converted to PLANET, HR 8799 f absolute magnitudes by adopting the magnitude of HR 8799 b from Metchev et al. (2009). For the LMIRCam 3.3 μm data, While HR 8799 is known to have four giant planets at wide we calculated photometry for all four planets with respect to separations, the inner system might have one or more com- unsaturated images of the star, obtained immediately after the panions that have not yet been discovered. Additional compan- saturated images used to detect the planets. We converted these ions would challenge formation models, which already have a to absolute photometry using HR 8799’s absolute magnitude at hard time explaining the outer four planets (Dodson-Robinson 3.3 μm from Hinz et al. (2010). Errors on the LMIRCam abso- et al. 2009; Kratter et al. 2010). A fifth planet would also compli- lute calibration are primarily the result of changing Strehl ratios cate dynamical stability analyses, which require mean-motion and telluric absorption variation throughout our observations. To resonances and planet masses that are on the low end of the test the magnitude of the Strehl-ratio variation, we performed range predicted by evolutionary models and independent age r = λ/D aperture photometry on the unsaturated data and found estimates (Fabrycky & Murray-Clay 2010; Moro-Mart´ın et al. a standard deviation of only ∼2%. To test the magnitude of the 2010; Sudol & Haghighipour 2012). Hinkley et al. (2011)used telluric absorption variation, we compared the Airy pattern of non-redundant masking to rule out the presence of massive inner ∼ the saturated and unsaturated images and found them to be con- companions from 0.01 to 0.5, but only for objects significantly sistent within ∼5%. Combining these error terms, we adopt an more massive than the four known planets. Here we use our absolute calibration error of 0.06 mag for the LMIRCam 3.3 μm H-band image to search for planetary mass companions interior data. H-band and 3.3 μm photometry of HR 8799 is presented to HR 8799 e. in Table 1. We evaluate our ability to detect a close-in companion by Our H-band photometry for HR 8799 c and d is consistent making a contrast curve of our residual H-band image (after the 15 with the results of Marois et al. (2008) and Metchev et al. (2009), four planets have been removed). We smooth the image with and we detect HR 8799 e for the first time at H band, finding it to a Gaussian that is the same size as our diffraction-limited PSF be ∼0.3 mag brighter than the next brightest planet (HR 8799 c). and calculate the standard deviation in 1 pixel annuli. Counts Our 3.3 μm data are somewhat inconsistent with previous are converted to photometry using the peak-flux (central pixel) photometry from Hinz et al. (2010) and Currie et al. (2011), of planet “b” from the smoothed image. We also correct for which are independent reductions of observations taken with self-subtraction, which is measured by inserting fake planets MMT/Clio (note that the Clio and LMIRCam 3.3 μm filters are into the raw data at various radii. Our 5σ contrast curve is identical). Hinz et al. (2010) reported detections of HR 8799 c shown in Figure 3. HR 8799 b–e are shown as diamonds. and d but not b, and Currie et al. (2011) reported detections of The vertical dashed line denotes the separation of the 2:1 HR 8799 b and c but not d. The most substantial disparity is for mean-motion resonance with HR 8799 e (assuming a face- HR 8799 b, for which Hinz et al. (2010) reported an absolute on circular orbit for simplicity). If there is a massive interior magnitude upper limit of 14.82, while Currie et al. (2011) planet, it is likely to be in a stable resonance, as has been reported a detection of 13.96 ± 0.28. Our detection of 13.22 found for pairs of the outer four planets. We find no fifth ± 0.11 is closer to the result of Currie et al. (2011) but is still planet at or exterior to the 2:1 resonance with HR 8799 e, with brighter by a significant amount. We reanalyzed the final reduced limits down to the approximate brightnesses (and by extension, masses) of the inner three planets. As a check, we insert a fake images from Hinz et al. (2010) and find the upper limit reported by Hinz et al. (2010) to be erroneous (likely a typographic error). “e”-like planet into our raw images at a separation of 0.235, the Our photometry for HR 8799 c is also brighter than observed by approximate position of the 2:1 resonance with “e.” Our reduced Hinz et al. (2010) and Currie et al. (2011), and our photometry image (Figure 4) shows that we would have detected “planet f” of HR 8799 d is brighter than observed by Hinz et al. (2010). anywhere exterior to the 2:1 resonance, and that there are no We present a comparison of the MMT/Clio image and our new residuals brighter than “f” in the image. Note that we do not LBT/LMIRCam image in Figure 2. The fact that our photometry repeat this analysis with the 3.3 μm image because it is not as is somewhat inconsistent with the values of Hinz et al. (2010) sensitive to additional companions as the H-band image. and Currie et al. (2011) can likely be explained by the non- photometric conditions reported in Hinz et al. (2010), varying AO performance (caused by the non-photometric conditions) 15 Note that we use an H-band image constructed with Nδ = 0.5 instead of and overly optimistic systematic and/or measurement error Nδ = 1.0 (which was described in Section 3) to maximize signal-to-noise ratio, as described in Lafreniere` et al. (2007). The Nδ = 0.5 contrast curve is analyses considering the quality of the Clio data. Variability ∼0.1–0.2 mag better than the Nδ = 1.0 contrast curve from ∼0. 2to0. 3and is unlikely to be a factor, given that it has not been reported then becomes progressively worse (up to ∼0.5 mag) outside of 0. 3. For this in any other photometric band and the amplitude of variability section, the inner regions are more important, so we use the Nδ = 0.5 contrast needed to rectify the disparity is quite large. curve.

4 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Figure 2. Comparison of the MMTAO/Clio 3.3 μm image from Hinz et al. (2010) and our new LBTAO/LMIRCam 3.3 μm image. Our 3.3 μm photometry is somewhat inconsistent with the ﬁndings of Hinz et al. (2010) and Currie et al. (2011), who separately analyzed the MMT data. Based on the relative quality of the images, it is likely that the disparity is the result of overly optimistic error bars by Hinz et al. (2010) and Currie et al. (2011). (A color version of this ﬁgure is available in the online journal.)

Figure 4. Image of the HR 8799 system with a fake planet, “f,” added in at the Figure 3. H-band contrast curve for the LBTAO/PISCES H-band image of approximate location of HR 8799 e’s 2:1 orbital resonance. The fake planet, HR 8799, with the four planets shown. Also shown is the position of the 2:1 which is the same brightness as HR 8799 e, was added into our individual raw orbital resonance with HR 8799 e (assuming, for simplicity, a face-on, non- frames and is easily recovered by our LOCI pipeline. There are no residual point eccentric orbit). If there is a massive inner planet, it is likely to be in a stable sources as bright as the fake planet at or exterior to its position at the HR 8799 resonance, as has been found for the outer companions. Based on the contrast e 2:1 orbital resonance. curve, we would have been able to detect a planet at HR 8799 e’s 2:1 inner (A color version of this figure is available in the online journal.) resonance, if it were approximately as bright (massive) as HR 8799 cde. Note that the contrast curve shows a dark hole inside of ∼0. 6, which is a predicted feature of high-order adaptive optics systems (Malbet et al. 1995). 2003) and DUSTY (Chabrier et al. 2000) atmospheric models, which the authors interpreted as evidence for non-equilibrium 5. MULTIWAVELENGTH MODELING OF THE chemistry (Janson et al. 2010). Three-color photometry in the HR 8799 PLANETS L and M bands was also inconsistent with equilibrium chemistry atmospheric models and, in particular, showed a relative Multiwavelength photometry and spectroscopy have im- lack of methane absorption at 3.3 μm (Hinz et al. 2010). H and proved our understanding of the physical properties of the K spectroscopy of HR 8799 b also showed a methane defi- HR 8799 planets, which are not well fit by the same models ciency (Bowler et al. 2010;Barmanetal.2011a). Currie et al. that have been used to interpret the properties of field brown (2011) and Madhusudhan et al. (2011) were able to parameter- dwarfs. Marois et al. (2008) first noted that the HR 8799 planet ize thick-cloud models to fit multiwavelength photometry for spectral energy distributions (SEDs) were best fit by model at- HR 8799 bcd. However, the models were not able to repro- mospheres with T = 1400–1700 K, while their luminosities duce the 3.3 μm photometry or the subsequent spectroscopy of were more consistent with T = 800–1000 K. The disparity was Barman et al. (2011a), who fit both the photometry and spec- driven by the planets’ faint/red appearance with respect to mod- troscopy of HR 8799 b with models that incorporated clouds els and a lack of methane absorption in narrowband 1.59/1.68 and non-equilibrium chemistry. CH4 photometry. Subsequent 3.88–4.10 μm spectroscopy of The combined result of these studies is that HR 8799 b, HR 8799 c was inconsistent with both COND (Baraffe et al. c, and d have non-equilibrium CO↔CH4 chemistry and cloudy

5 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Figure 5. Left: H vs. H − K color–magnitude diagram showing the M→L→T spectral-type transition for field brown dwarfs (Leggett et al. 2002; Knapp et al. 2004) and the HR 8799 planets. The HR 8799 planets appear to be an extension of the L-dwarf sequence, implying that they have cloudy atmospheres at lower effective temperatures than are typical for cloudy field brown dwarfs. Right: L vs. 3.3 μm−L color–magnitude diagram showing equilibrium chemistry, thick-cloud atmospheres from Madhusudhan et al. (2011), and the HR 8799 planets. The HR 8799 planets are all brighter at 3.3 μm than predicted by the Madhusudhan et al. (2011) models, implying a lack of CH4, which is a strong absorber at 3.3 μm in the equilibrium chemistry model atmospheres. (A color version of this figure is available in the online journal.)

Table 2 Photometry of the HR 8799 Planets

Planet zJHCH4sCH4l Ks 3.3 L M (1.03 μm) (1.25 μm) (1.63 μm) (1.59 μm) (1.68 μm) (2.15 μm) (3.3 μm) (3.8 μm) (4.7 μm) HR 8799 b 18.24 ± 0.29 16.30 ± 0.16 15.08 ± 0.13 15.18 ± 0.17 14.89 ± 0.18 14.05 ± 0.08 13.2 ± 0.11 12.66 ± 0.11 13.07 ± 0.30 HR 8799 c 14.65 ± 0.17 14.18 ± 0.14 14.25 ± 0.19 13.90 ± 0.19 13.13 ± 0.08 12.2 ± 0.11 11.74 ± 0.09 12.05 ± 0.14 HR 8799 d 15.26 ± 0.43 14.23 ± 0.2 14.03 ± 0.30 14.57 ± 0.23 13.11 ± 0.12 12.0 ± 0.11 11.56 ± 0.16 11.67 ± 0.35 HR 8799 e 13.88 ± 0.2 12.93 ± 0.22 12.1 ± 0.21 11.61 ± 0.12 Reference (b,c,d,e) 2 3,3,3 4,1,1,1 3,3,3 3,3,3 3,3,3,5 1,1,1,1 3,3,3,5 6,6,6

References. (1) This work; (2) Currie et al. 2011; (3) Marois et al. 2008; (4) Metchev et al. 2009; (5) Marois et al. 2010; (6) Galicher et al. 2011. atmospheres at low effective temperatures where, in field brown Table 3 dwarfs, the clouds are thought to have settled below the pho- Colors of the HR 8799 Planets tosphere. Two color–magnitude diagrams, shown in Figure 5, Planet J − H H − Ks Ks − L 3.3 μm−L L − M − demonstrate these effects. On the left is an H K versus H ± ± ± ± − ± → → HR 8799 b 1.22 0.20 1.03 0.15 1.39 0.13 0.54 0.15 0.41 0.31 color–magnitude diagram, which shows the M L Tse- HR 8799 c 0.47 ± 0.21 1.05 ± 0.16 1.39 ± 0.12 0.46 ± 0.14 −0.31 ± 0.16 quence of field brown dwarfs. L dwarfs are characterized by HR 8799 d 1.03 ± 0.46 1.12 ± 0.23 1.55 ± 0.20 0.44 ± 0.19 −0.11 ± 0.37 cloudy atmospheres, while T dwarfs are characterized by cloud- HR 8799 e 0.95 ± 0.29 1.32 ± 0.25 0.49 ± 0.24 free atmospheres. The intermediate region is the L→T transition, where objects are thought to have patchy clouds or clouds that have partially descended below the photosphere. all model comparisons in this paper, we convolve the model The HR 8799 planets appear to be an extension of the field planet atmosphere and a model of Vega (Cohen et al. 1992) with L-dwarf sequence, i.e., they have clouds at an effective temper- filter profiles to produce predicted magnitudes in the different ature where the field brown dwarfs are transitioning to cloudless. filters, which are then compared to the measured photometry. The right-hand side of Figure 5 shows a 3.3 μm−L versus L The filter profiles have all been multiplied by a model telluric color–magnitude diagram, which includes a sequence of chem- atmosphere.16 For most filters, this step has a negligible effect, ical equilibrium, thick-cloud models from Madhusudhan et al. because the telluric atmosphere has a flat transmission profile, (2011). The HR 8799 planets are all much brighter at 3.3 μm but Earth’s atmosphere has a large transmission slope in the than predicted by the models, implying a relative absence of 3.3 μm filter that changes the effective wavelength of the filter CH4, which is a strong absorber at 3.3 μm. and, in turn, changes the model magnitudes by ∼0.1 mag. For the In the following sections, we present model atmospheres sake of plotting the model fits, we convert the filter magnitudes of the HR 8799 planets in an attempt to reproduce their to Jy by convolving the Vega model with filter profiles to photometry. A summary of the available photometry is presented in Table 2, colors are presented in Table 3, and a listing of 16 http://www.gemini.edu/sciops/ObsProcess/obsConstraints/ all the models used in this paper is presented in Table 4.For atm-models/cptrans_zm_43_15.dat

6 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal. Table 4 with interior evolutionary models, which predict a larger object Atmospheric Models Used in This Paper radius. Barman et al. (2011a) also present model atmospheres that obey the predictions of the evolutionary tracks, with a best- Figure No. Teff (K) log(g) Cloud Type Chemistry fit model that has T = 896 K and Z = 1.0 metallicity. This 6 850 4.3 AE.60 Equilibrium model demonstrates that it might be possible to explain HR 8799 10 ×CO, 0.1×CH4 100×CO, 0.01×CH b’s appearance with a combination of clouds, non-equilibrium 4 chemistry, and higher than solar metallicity. However, the model 7 1400 4.0 AE.60 Equilibrium does not fit all of the existing photometry, and our new 3.3 μm 700 A.100 photometry makes the fit significantly worse. 8 1000 4.2 AE.60 Equilibrium ↔ 10 ×CO, 0.1×CH4 5.1.1. Non-equilibrium CO CH4 Chemistry 100×CO, 0.01×CH4 900 3.8 AE.60 Equilibrium We start our modeling by using the parameterized thick-cloud 10 ×CO, 0.1×CH4 atmospheres of Madhusudhan et al. (2011) and adjusting the 100×CO, 0.01×CH4 CO and CH4 mixing ratios. These models assume evolutionary track radii but make no attempt to self-consistently explain 9 1400 4.0 AE.60 Equilibrium 700 A.100 the CO and CH4 mixing ratios, which Barman et al. (2011a) produce with turbulent mixing. In Figure 6, we plot the best-fit 10 1000 4.0 AE.60 Equilibrium chemical equilibrium model of HR 8799 b from Madhusudhan 11 1000 4.0 AE.60 Equilibrium et al. (2011), as well as two models that have suppressed CH4 × × 10 CO, 0.1 CH4 and enhanced CO with respect to the equilibrium models (by 100×CO, 0.01×CH 4 factors of 10 and 100, respectively). Suppressing CH4 and 12 1400 4.0 AE.60 Equilibrium enhancing CO has a negligible effect on the near-infrared (zJHK) 700 A.100 photometry, but it greatly affects the HK spectroscopy, favoring the 100×CO, 0.01×CH4 model. Non-equilibrium chemistry Notes. Cloud types (“A” and “AE”) refer to cloud thickness, and the associated dramatically affects the 3–5 μm SED, where a lack of CH4 numbers (“60” and “100”) refer to modal dust grain size, as parameterized makes the object brighter in the 3.3 μm and L filters and excess in Madhusudhan et al. (2011). Multiples of CO and CH4 are with respect to CO makes the object fainter in the M-band filter. The 100×CO, equilibrium chemistry models. All models assume solar metallicity. Further 0.01×CH4 model fits the 3.3 μm photometry and comes closest model details are discussed in Burrows et al. (2006) and Madhusudhan et al. M (2011). to fitting the -band photometry (further increasing the CO mixing ratio would improve this fit). However, the lack of CH4 strongly increases the predicted flux in the L filter, making it produce zero-point fluxes. For HR 8799 b, we also include incompatible with HR 8799 b’s observed L photometry. Based the H and K spectroscopy from Barman et al. (2011a). The HK on this analysis, it seems unlikely that a cool (850 K) atmosphere spectrum comparison is made by smoothing the planet model with non-equilibrium chemistry could explain HR 8799 b’s atmosphere by the published spectrum’s resolution (0.01 μm) 3.3 μm−L color. However, there remains the possibility that and directly comparing to the observed spectrum. The absolute more complex radial profiles of non-equilibrium chemistry and fluxes of the H and K spectroscopy are tied to the H and clouds could explain all of the data. K photometry, so for our comparison we allow the overall brightness of the two spectra to vary and only fit the shapes of the 5.1.2. Mixed-cloud Atmospheres spectra. Based on Figure 5, HR 8799 b has colors reflective of ∼ 5.1. HR 8799 b a 1300 K atmosphere, but a luminosity consistent with an ∼850 K atmosphere. In lieu of resolving the difference HR 8799 b is the most challenging atmosphere to explain by assuming a small (unphysical) radius, we consider the because it is the coolest of the four planets and is thus the largest possibility that the planet emits non-isotropically, with bright outlier in the color–magnitude diagrams shown in Figure 5. and dark sections, such that the bright sections dominate the Additionally, the HR 8799 b atmosphere is the most constrained shape of the SED. Bright and dark regions have been observed of the four planets due to the H and K spectroscopy of Barman on Jupiter in the mid-infrared (Westphal 1969), and several et al. (2011a). Our new 3.3 μm photometry of HR 8799 b is studies seeking to explain the L→T transition have used significantly brighter (by ∼100%) than the previously published hybrid cloudy/cloud-free models (e.g., Burgasser et al. 2002). valuebyCurrieetal.(2011). As described in Section 3,the For our purposes, mixtures of standard cloudy and cloud-free Currie et al. (2011)3.3μm photometry is likely erroneous, models are unlikely to explain the appearance of HR 8799 b due to non-photometric conditions and/ory overly optimistic because its H − K color is redder than the L-dwarf sequence, error bars. There is the possibility that HR 8799 b is extremely implying that it is more cloudy, not less cloudy, than the other variable at 3.3 μm, but we consider that scenario unlikely L dwarfs. due to the many times HR 8799 has been observed at other An alternative is that HR 8799 b has a mixture of clouds, wavelengths where there has been no evidence of such large such that the whole planet is cloudy, but with regions that variability. have thicker clouds where the planet appears darker. A coarse Before proceeding with our modeling, we examine Barman approximation of this phenomenon is to linearly combine cloudy et al.’s (2011a) recent hypothesis regarding the SED of HR models of different effective temperatures (analogously to how 8799 b. Barman et al. (2011a) were able to reproduce all Burgasser et al. 2002 and others have linearly combined existing photometry and spectroscopy of HR 8799 b by using cloudy and cloud-free models of different effective tempera- a non-equilibrium CO↔CH4 cloudy atmosphere with Teff = tures). This method is somewhat non-physical due to the fact 1100 K and log(g) = 3.5. However, their model is inconsistent that the two atmospheres have different temperature–pressure

7 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Figure 6. Photometry and spectroscopy of HR 8799 b (red error bars; Table 2 and Barman et al. 2011a) with atmospheric models (green, pink, and blue curves) and predicted in-band fluxes for each photometry point (green, pink, and blue horizontal lines, which span the filters’ half-max profiles). The 850 K equilibrium chemistry model (green) is the best-fit “thick-cloud” model for HR 8799 b from Madhusudhan et al. (2011). Two other models (pink and blue) suppress the CH4 mixing ratios and enhance the CO mixing ratios by 10× and 100× with respect to the equilibrium chemistry model. The H and K spectroscopy are well fit by the non-equilibrium chemistry models. Our new 3.3 μm photometry is best fit by the 100×CO, 0.01×CH4 model. However, this model predicts a flux that is substantially higher than published measurements at 3.8 μm(L). None of the models are able to reproduce the relatively flat SED from 3.3 to 3.8 μm. (A color version of this figure is available in the online journal.)

Figure 7. Same as Figure 6 but with a mixed-cloud model (93% 700 K A-type clouds and 7% 1400 K AE-type clouds from Madhusudhan et al. 2011). The fit adequately reproduces all photometry (except in the M-band filter, where additional CO absorption would rectify the discrepancy) and greatly flattens the SED from 3.3 to 3.8 μm compared to the non-equilibrium models shown in Figure 6.AtH and K, the model generally reproduces the shape of HR 8799 b’s observed spectrum (i.e., negligible CH4 absorption) but is flatter. (A color version of this figure is available in the online journal.) profiles (Marley et al. 2010). Linearly combining models with paper. In Figure 7, we present an example of a hybrid atmo- a shared temperature–pressure profile but different cloud struc- sphere that is 93% Teff = 700 K, “A”-type cloudy and 7% tures would be more correct but is beyond the scope of this Teff = 1400 K, “AE”-type cloudy (A and AE cloud profiles

8 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Figure 8. Same as Figure 6 (top) but for HR 8799 c and d. As was found for HR 8799 b, our non-equilibrium chemistry models are unable to ﬁt the 3.3 μm−L colors of HR 8799 c and d. (A color version of this ﬁgure is available in the online journal.)

Figure 9. Same as Figure 7 (top) but for HR 8799 c and d. We purposely choose the same cloudy atmospheres to mix as were used for HR 8799 b. We are able to fit HR 8799 c and d with mixed-cloud atmospheres, but using a higher mixing fraction of the 1400 K atmosphere than was found for HR 8799 b. The addition of non-equilibrium chemistry to these models would likely improve the fit. (A color version of this figure is available in the online journal.) are described in Madhusudhan et al. (2011), and further model but its bulk shape is generally correct (i.e., it does not show details are presented in Burrows et al. 2006). The hybrid strong CH4 absorption, as would be expected for a cool ob- model adequately fits all of the photometry except for M band, ject). We note that this model is meant to be representative, which can be explained with further increased CO absorp- but that a true mixed-cloud atmosphere should be calculated tion. The HK model spectrum is muted compared to the data, self-consistently.

9 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Figure 10. Thick-cloud atmosphere for HR 8799 e (plotted with the same symbols used in Figures 6–9). In this fit, we ignore the 3.3 μm photometry as was done by Madhusudhan et al. (2011) when fitting HR 8799 b, c, and d. (A color version of this figure is available in the online journal.)

Figure 11. Same as Figures 6 (top) and 8 but for HR 8799 c and d. As was found for HR 8799 b, c, and d, our non-equilibrium chemistry models are unable to ﬁt the 3.3 μm−L colors of HR 8799 e. (A color version of this ﬁgure is available in the online journal.)

5.2. HR 8799 c, d, and e available photometry with non-equilibrium chemistry models, and a mixed-cloud model can be made to fit. We repeat our analysis of HR 8799 b (Section 5.1)for HR 8799 c and d, again starting with the best-fit, thick-cloud 5.3. The HR 8799 Planets in Aggregate models from Madhusudhan et al. (2011). A comparison of The HR 8799 system provides a unique laboratory for non-equilibrium CO↔CH models is shown in Figure 8, and 4 simultaneously studying multiple coeval planets. In the previous example mixed-cloud atmospheres are shown in Figure 9.Our sections, we have modeled the planets individually. Comparing conclusions for HR 8799 c and d are similar to our conclusions the four planets provides additional insight. for HR 8799 b: we are unable to fit the 3.3 μm−L colors of HR HR 8799 c, d, and e are brighter than HR 8799 b in all 8799 c and d with cloudy/non-equilibrium chemistry models, published photometry but have almost exactly the same colors and mixed-cloud atmospheres do a reasonable job fitting all of as “b” throughout their measured SEDs (see Table 3; the colors the data. We purposely construct the mixed-cloud atmospheres for all four planets are the same, within errors, except for J − H from the same two model atmospheres used to make the mixture for HR 8799 c, where J is abnormally bright). Given that the for HR 8799 b (see Section 5.1.2). In this scenario, we find that planets are coeval, they are expected to all have similar radii (to HR 8799 c and d have higher fractions of the warm atmosphere within ∼10%; Burrows et al. 1997). Therefore, the fact that “c,” than HR 8799 b, explaining their higher luminosity. “d,” and “e” are brighter than “b” implies that they have higher HR 8799 e has less data than the outer three planets, but effective temperatures (based on L = 4πR2σT4 ). However, the the existing data are consistent with the photometry of HR eff fact that all of the HR 8799 planets have the same colors suggests 8799 c and d. With the addition of our new H-band and that the physical properties of their atmospheres are similar, 3.3 μm photometry, HR 8799 e has now been studied at four despite their different effective temperatures. In field brown wavelengths, and we can proceed with atmospheric modeling, dwarfs, objects with different effective temperatures (over the based on lessons learned from HR 8799 b, c, and d. We begin range probed by the HR 8799 planets) have different physical by using the thick-cloud models from Madhusudhan et al. properties and different SED colors. For the HR 8799 planets, (2011) to look for a good fit of the H, K, and L photometry. this does not appear to be the case.17 We find that a 1000 K, log(g) = 4.0 model fits well (shown The fact that “c,” “d,” and “e” have almost the exact same in Figure 10). Based on the small number of data points, colors as “b” is circumstantial evidence for the mixed-cloud degeneracies between different cloud properties, surface gravity, atmospheres. In the mixed-cloud scenario, HR 8799 b has a and effective temperature are not explored, but it is reasonable lower effective temperature than the other HR 8799 planets, but to assume that the cloud properties and surface gravity will be similar to the other HR 8799 planets. Figures 11 and 12 show 17 As an example, Dupuy & Liu (2012)findΔK − L /ΔK = 0.4 over a large non-equilibrium chemistry models and mixed-cloud models. effective temperature range, whereas HR 8799 b has a ΔK − L/ΔK = 0.03 As was true for the outer three planets, we are unable to fit the with respect to the average values of HR 8799 c, d, and e.

10 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Figure 12. Same as Figures 7 (top) and 9 but for HR 8799 e. The mixing fraction of the two atmospheres is similar to what we found for HR 8799 c and d. The addition of non-equilibrium chemistry to these models would slightly improve the fit, although the photometric error bars for HR 8799 e are large enough that this is not necessary. (A color version of this figure is available in the online journal.) a similar physical temperature in the bright, emitting regions by the HR 8799 planets’ bolometric magnitudes. As a result, we of its atmosphere. In Figures 7, 9, and 12, we have shown that consider the possibility that small sections of the planets’ atmo- all of the HR 8799 planets can be fit by a mixture of 1400 K spheres are hot (>1300 K), dominating the shape of the SEDs, and 700 K cloudy atmospheres (which we have been using while the majority of the planets’ atmospheres are cooler and do as an approximation for two different cloud structures, one of not produce much flux. The temperature–pressure profile must which has much lower opacity). The difference between the be the same between the “hot” and “cool” regions, so the phys- cooler planet, “b,” and the hotter planets, “c,” “d,” and “e,” is the ical difference would be that the “cool” regions have increased mixing ratio of the 1400 K and 700 K models, i.e., the covering cloud opacity. Since the SED is consistent with a cloudy atmo- fraction of the different cloud types. sphere, the “hot” regions must also be cloudy, so the combined The non-equilibrium chemistry models (with reasonable atmosphere is composed of mixed clouds, some of which are radii) of Barman et al. (2011a)fittheHK spectroscopy bet- thicker than others. Our mixed-cloud models are able to fit all of ter than our makeshift mixed-cloud models, but they do not the HR 8799 data. However, we caution that our models are not reproduce the broadband colors as well, in particular from 3.3 fully self-consistent and that more theoretical work is necessary to 3.8 μm. The Barman et al. (2011a) models do provide physi- to validate our hypothesis. cal motivations for non-equilibrium chemistry (turbulent radial The HR 8799 planets have unusual SEDs that are not well mixing), while our mixed-cloud models are based on analogy fit by the same models that have been used to fit field brown with physically quite different systems (Jupiter and L/T transi- dwarfs. From an observational standpoint, HK spectroscopy and tion brown dwarfs). Self-consistent modeling is still necessary 3.3 μm−L colors have been particularly powerful in ruling to determine if mixed-cloud atmospheres are a viable explana- out model atmospheres. HK spectroscopy of HR 8799 cde, tion for the HR 8799 planets. In any case, it appears that detailed and other directly imaged planets in general, is critical to our modeling of increasingly complex cloud physics and chemistry understanding of clouds and non-equilibrium chemistry. We also will be necessary to explain the true nature of the HR 8799 note that, given the wide range of 3–4 μm SEDs predicted by planets. the models in this paper, it would be very useful to obtain low- resolution spectroscopy of the HR 8799 planets in this range. 6. SUMMARY AND CONCLUSIONS From a theoretical standpoint, our new 3.3 μm photometry is We have directly imaged the HR 8799 planetary system, a challenge even for non-equilibrium chemistry models, which μ detecting all four planets at H band and 3.3 μm with the predict bright 3.3 m photometry. Mixed-cloud models are one μ μ LBT’s FLAO system. The images are of unprecedented quality, possible way to flatten out 3.3 m–3.8 m photometry and hide allowing us to rule out the presence of a massive (HR 8799 cde- CH4 opacity. like) planet exterior to HR 8799 e’s 2:1 inner resonance (H band 5σ contrast of 11.6 mag at 0.235). We detect HR 8799 e at The authors thank Piero Salinari for his insight, leadership, H band, for the first time, and find that it is approximately as and persistence, which made the development of the LBT adaptive secondaries possible. We also thank the many dedicated bright as HR 8799 c and d. Combined with Ks and L data (Marois et al. 2010) and our new 3.3 μm data, this indicates that individuals who have worked on LBTI, LMIRCam, PISCES, HR 8799 e has similar atmospheric properties to HR 8799 c and and the AO system over the years. The authors thank the anony- d. The planets are all brighter than expected at 3.3 μm, where mous referee for his/her helpful report. The Large Binocular Telescope Interferometer is funded by the National Aeronautics equilibrium chemistry models predict that CH4 opacity should make the planets faint. and Space Administration as part of its Exoplanet Exploration We model the HR 8799 planets with thick-cloudy atmo- program. LMIRCam is funded by the National Science Foun- dation through grant NSF AST-0705296. spheres (Madhusudhan et al. 2011) and allow the CO and CH4 mixing ratios to vary arbitrarily. The models that fit our 3.3 μm data (which have very little CH4) substantially overpredict the REFERENCES planets’ fluxes at L (3.8 μm). Hotter atmospheres (>1300 K) have a similar 3.3 μm–3.8 μm color as the HR 8799 planets, and Baraffe, I., Chabrier, G., Barman, T. S., Allard, F., & Hauschildt, P. H. 2003, A&A, 402, 701 the shape of the Barman et al. (2011a) HK spectrum is also well Barman, T. S., Macintosh, B., Konopacky, Q. M., & Marois, C. 2011a, ApJ, fit by an atmosphere that is significantly hotter than indicated 733, 65

11 The Astrophysical Journal, 753:14 (12pp), 2012 July 1 Skemeretal.

Barman, T. S., Macintosh, B., Konopacky, Q. M., & Marois, C. 2011b, ApJ, Hinz, P. M., Rodigas, T. J., Kenworthy, M. A., et al. 2010, ApJ, 716, 417 735, L39 Janson, M., Bergfors, C., Goto, M., Brandner, W., & Lafreniere,` D. 2010, ApJ, Bowler, B. P., Liu, M. C., Dupuy, T. J., & Cushing, M. C. 2010, ApJ, 723, 710, L35 850 Knapp, G. R., Leggett, S. K., Fan, X., et al. 2004, AJ, 127, 3553 Burgasser, A. J., Marley, M. S., Ackerman, A. S., et al. 2002, ApJ, 571, L151 Kratter, K. M., Murray-Clay, R. A., & Youdin, A. N. 2010, ApJ, 710, 1375 Burrows, A., Marley, M., Hubbard, W. B., et al. 1997, ApJ, 491, 856 Lafreniere,` D., Marois, C., Doyon, R., Nadeau, D., & Artigau, E.´ 2007, ApJ, Burrows, A., Sudarsky, D., & Hubeny, I. 2006, ApJ, 640, 1063 660, 770 Burrows, A., Sudarsky, D., & Lunine, J. I. 2003, ApJ, 596, 587 Leggett, S. K., Golimowski, D. A., Fan, X., et al. 2002, ApJ, 564, 452 Chabrier, G., Baraffe, I., Allard, F., & Hauschildt, P. 2000, ApJ, 542, 464 Lloyd-Hart, M. 2000, PASP, 112, 264 Chauvin, G., Lagrange, A.-M., Dumas, C., et al. 2004, A&A, 425, L29 Madhusudhan, N., Burrows, A., & Currie, T. 2011, ApJ, 737, 34 Cohen, M., Walker, R. G., Barlow, M. J., & Deacon, J. R. 1992, AJ, 104, 1650 Malbet, F., Yu, J. W., & Shao, M. 1995, PASP, 107, 386 Cowley, A., Cowley, C., Jaschek, M., & Jaschek, C. 1969, AJ, 74, 375 Marley, M. S., Saumon, D., & Goldblatt, C. 2010, ApJ, 723, L117 Currie, T., Burrows, A., Itoh, Y., et al. 2011, ApJ, 729, 128 Marois, C., Lafreniere,` D., Doyon, R., Macintosh, B., & Nadeau, D. 2006, ApJ, Dodson-Robinson, S. E., Veras, D., Ford, E. B., & Beichman, C. A. 2009, ApJ, 641, 556 707, 79 Marois, C., Macintosh, B., Barman, T., et al. 2008, Science, 322, 1348 Dupuy, T. J., & Liu, M. C. 2012, arXiv:1201.2465 Marois, C., Zuckerman, B., Konopacky, Q. M., Macintosh, B., & Barman, T. Esposito, S., Mesa, D., Skemer, A., et al. 2012, arXiv:1203.2735 2010, Nature, 468, 1080 Esposito, S., Riccardi, A., Pinna, E., et al. 2011, Proc. SPIE, 8149, 814902 McCarthy, D. W., Jr., Ge, J., Hinz, J. L., Finn, R. A., & de Jong, R. S. 2001, PASP, Esposito, S., Riccardi, A., Quiros-Pacheco,´ F., et al. 2010, Appl. Opt., 49, G174 113, 353 Fabrycky, D. C., & Murray-Clay, R. A. 2010, ApJ, 710, 1408 Metchev, S., Marois, C., & Zuckerman, B. 2009, ApJ, 705, L204 Galicher, R., Marois, C., Macintosh, B., Barman, T., & Konopacky, Q. Moor,´ A., Abrah´ am,´ P., Derekas, A., et al. 2006, ApJ, 644, 525 2011, ApJ, 739, L41 Moro-Mart´ın, A., Rieke, G. H., & Su, K. Y. L. 2010, ApJ, 721, L199 Geballe, T. R., Knapp, G. R., Leggett, S. K., et al. 2002, ApJ, 564, 466 Moya, A., Amado, P. J., Barrado, D., et al. 2010, MNRAS, 405, L81 Gray, R. O., & Kaye, A. B. 1999, AJ, 118, 2993 Saumon, D., & Marley, M. S. 2008, ApJ, 689, 1327 Heinze, A. N., Hinz, P. M., Kenworthy, M., Miller, D., & Sivanandam, S. Skemer, A. J., Close, L. M., Szucs,˝ L., et al. 2011, ApJ, 732, 107 2008, ApJ, 688, 583 Skrutskie, M. F., Jones, T., Hinz, P., et al. 2010, Proc. SPIE, 7735, 77353H Hinkley, S., Carpenter, J. M., Ireland, M. J., & Kraus, A. L. 2011, ApJ, 730, Su, K. Y. L., Rieke, G. H., Stapelfeldt, K. R., et al. 2009, ApJ, 705, 314 L21 Sudol, J. J., & Haghighipour, N. 2012, arXiv:1201.0561 Hinz, P. M., Bippert-Plymate, T., Breuninger, A., et al. 2008, Proc. SPIE, 7013, Westphal, J. A. 1969, ApJ, 157, L63 701328 Zuckerman, B., Rhee, J. H., Song, I., & Bessell, M. S. 2011, ApJ, 732, 61